An Error Allocation Optimization Design Method for a Visual Simulation System

Abstract

The present invention relates to an error allocation optimization design method for a visual simulation system. Based on the radiation energy transfer process of the simulation system, error influencing factors are comprehensively analyzed, and a radiation energy error transfer function of the simulation system is constructed based on the error transfer theory. For the main error factors, an objective function for error optimization allocation of the simulation system is constructed. The adjustment range of each error factor is set through the feasibility of technical implementation. An experiment is designed. Taking the error transfer function as the optimization constraint function, the particle swarm optimization algorithm is used to optimize the objective function within the adjustment range of each error factor, and the optimal error allocation scheme of the simulation system is output. The present invention ensures that the error of the visual simulation system is controlled within a certain range, realizes the minimum technical implementation difficulty of the visual simulation system, and obtains the best economic benefits.

Description

Technical Field

[0001] The present invention relates to the technical field of performance analysis of a simulation system, and more specifically to an error allocation optimization design method for a visual simulation system. Background Art

[0002] Visual simulation technology has been applied by more and more countries in more and more fields. The visual simulation imaging technology has currently achieved rapid development. To test the imaging performance of a camera, there are generally methods such as mathematical simulation, outdoor testing, traditional indoor experiments, and hardware-in-the-loop simulation. Among them, the hardware-in-the-loop simulation combines actual objects with virtual digital scenes and introduces them into the test loop, which is more accurate than digital simulation. Therefore, in recent years, conducting hardware-in-the-loop simulation tests on the system in an indoor environment has gradually become a hot research topic in various countries. The advantage of using simulation technology is that it can eliminate the disadvantages of being greatly restricted by climatic conditions and consuming a large amount of manpower and material resources in outdoor tests, and reflect the camera imaging performance law in the real environment through a large number of indoor simulation test results.

[0003] However, a simulation system is not equivalent to the real environment. If an evaluation test is carried out using a simulation system that does not meet the standards, it will cause serious consequences of wrong conclusions. Through analysis and quantification, the errors generated by the simulation system can be accurately estimated, and the allocation optimization of the errors contributes to the further improvement of the simulation system. The feasibility and economy of system error optimization are the key issues for error allocation optimization.

[0004] The error allocation optimization of a simulation system has always been the focus and development direction of research in this field. In the field of system error optimization, many scholars and experts have carried out relevant research work. In the specific system error optimization allocation, Qian Huanan et al. from Shanghai Jiao Tong University, based on a press system, comprehensively considered the processing cost and the reliability of the mechanism output to optimize the allocation of the accuracy of key components, and then gave the measures that should be taken in the processing and assembly of each component to obtain the system error optimization allocation scheme. Wei Zongkang et al. from Beijing Aerospace Control Instrument Institute, based on an inertial measurement system, according to the overall accuracy requirement value and the proportion of each error source in the landing point accuracy, adaptively allocated each error coefficient, and finally achieved the optimal error coefficient index under the accuracy index requirements by successively adjusting the error term with the largest proportion in the landing point accuracy.

[0005] From the current technical development status, the methods for system error allocation optimization can be divided into the following categories:

[0006] (1) Based on the sensitivity analysis of error factors, the system is allocated errors. However, during the error allocation process, the sensitivity of the error factors will change to a certain extent. Therefore, it is not reasonable enough to allocate errors only based on the initial sensitivity.

[0007] (2) Evenly distribute the error to each error factor. This method simply considers the influence degree of all error factors to be the same, which may ultimately lead to too high system optimization costs.

[0008] (3) Set the error regulation range based on a certain error control feasibility, and then use an adaptive algorithm to select a suitable error distribution scheme. This method does not require sensitivity analysis of error factors for each system, is more popular, and can be adaptively adjusted according to different systems.

[0009] Although some achievements have been made in the field of system error distribution optimization, the adaptive optimization effect of error factors in the research results is poor, and the optimization scheme has a strong correlation with the actual system, making it impossible to apply it to the visual simulation system. How to establish an optimization method that can achieve the most economical error distribution optimization scheme within the total system error index and provide guidance for the further optimization design of the system. Such related methods are rarely used in the field of environmental simulation. Summary of the Invention

[0010] In view of the deficiencies of the prior art, the present invention provides an error distribution optimization method based on a visual simulation system. The optimization method first analyzes the system and constructs an energy error transfer model, and establishes a set of main error factors. Based on the energy error transfer model, and considering the difficulty of realizing the main technologies of the visual simulation system, a particle swarm algorithm is used to solve the system error distribution optimization scheme, providing a feasible scheme for the further optimization design of the visual simulation system.

[0011] The technical solution adopted by the present invention to achieve the above object is:

[0012] An error distribution optimization design method for a visual simulation system, comprising the following steps:

[0013] Construct an energy error transfer function based on the object energy of an object in an urban environment;

[0014] Construct an objective function based on error factors and determine the adjustment range of each error factor;

[0015] Use the particle swarm algorithm, take the energy error transfer function as a constraint function, and optimize the objective function within the adjustment range to obtain the error amount of each error factor as the optimal error distribution scheme.

[0016] The object energy L is:

[0017] L(θ,λ) = {εL T + [E S cos(θ S ) + Ld](1 - ε)}τ + Lu

[0018] Among them, ε is the emissivity of the object, and L T is the energy radiated by the object itself; τ is the transmittance of the observation path, and E S is the solar radiation, and θ S is the angle between the sun and the normal vector of the target surface; Ld is the downward atmospheric radiation, and Lu is the upward atmospheric radiation.

[0019] When the error factors are independent of each other, that is, a change in one error does not cause a change in another error, the energy error transfer function is:

[0020]

[0021] Among them, L is the object energy, θ is the error factor, Δθ is the error range;

[0022] When the error factors are correlated, that is, a change in one error causes a change in another error, the energy error transfer function is:

[0023] error 误差 =error L +xitem

[0024] Among them, xitem is the coupling term error.

[0025] The coupling term error is:

[0026]

[0027] Among them, L is the object energy, x i 、x j are the error factors, is the error range, is the correlation coefficient of the two errors.

[0028] The error influencing factors include: the self-thermal radiation of the target, the emissivity of the target, the transmittance of the observation path, the incident light angle of the target surface, the upward atmospheric radiation, the downward atmospheric radiation, and the focal length.

[0029] The objective function is:

[0030]

[0031] Among them, is the weight value of each component in the objective function, x(i)_change is the error amount after controlling the i-th error factor, and x(i)_error is the error calibration amount of the i-th error factor.

[0032] The specific particle swarm algorithm is:

[0033] The error factors are taken as particles and initialized as a group of random particles, flying at a set speed in a D-dimensional search space. When each particle is searching, when the optimal solution is found, the current position is replaced with the optimal solution and the search continues until the search ends.

[0034] The present invention has the following beneficial effects and advantages:

[0035] 1. The present invention uses the energy error transfer formula to analyze and quantify the errors of the visual simulation system, takes into account the correlation of error factors, adds the coupling term error to it, and ensures the reliability of error estimation.

[0036] 2. The present invention uses the particle swarm algorithm to calculate the error allocation optimization scheme of the system, ensures that in different systems, the error influencing factors and the cost control function are determined, and a relatively high-benefit error allocation scheme can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is the overall flowchart of the present invention;

[0038] Figure 2 is the schematic diagram of the energy transfer of the visual simulation system of the present invention;

[0039] Figure 3 is the flowchart of the particle swarm optimization of the present invention;

[0040] Figure 4a is the curve of the change of the error allocation optimization control process of the present invention - the optimization iteration process of the objective function;

[0041] Figure 4b is the curve of the change of the error allocation optimization control process of the present invention - the convergence process of the system error. DETAILED DESCRIPTION OF THE INVENTION

[0042] The following further describes the present invention in detail with reference to the drawings and embodiments.

[0043] The technical solution adopted by the present invention to achieve the above object is:

[0044] An error allocation optimization design method for a visual simulation system, characterized by including the following steps:

[0045] Step 1: Based on the radiation energy transfer process of the simulation system, comprehensively analyze the error influencing factors, and construct a radiation energy error transfer function of the simulation system based on the error transfer theory;

[0046] Step 2: Construct an objective function for the error optimization allocation of the simulation system for the main error factors;

[0047] Step 3: Determine the range of each error factor according to the technical feasibility of actually controlling each error factor;

[0048] Step 4: Design an experiment, use the error transfer function as the optimization constraint function, and use the particle swarm optimization algorithm to optimize the objective function within the adjustment range of each error factor, and output the optimal error allocation scheme of the simulation system.

[0049] The energy transfer process of the system is as follows:

[0050] By collecting the emissivity ε of the object, the energy L radiated by the object itself T , the path transmittance τ, the solar radiation E S , the angle θ between the sun and the normal vector of the target surface S , the downward atmospheric radiation Ld, the upward atmospheric radiation Lu, according to the object radiance calculation formula:

[0051] L(θ,λ) = {εL T + [E S cos(θ S ) + Ld](1 - ε)}τ + Lu

[0052] Obtain the digital apparent energy field, project it, and finally obtain an image through the camera.

[0053] The main error influencing factors include: the self-thermal radiation of the target, the emissivity of the target, the path transmittance of the observation path, the incident light angle of the target surface, the upward atmospheric radiation, the downward atmospheric radiation, and the focal length.

[0054] The construction of the error transfer theory is as follows:

[0055] If the error factors are independent of each other, that is, a change in one error will not cause a change in another error, then:

[0056]

[0057] Among them, L is the object energy, θ is the error factor, Δθ is the error range, error L is the error of the energy after being projected by the visual simulation system;

[0058] If there is a certain correlation between the error factors, then add the coupling term error on the premise of taking the error as independent:

[0059]

[0060] Among them, L is the object energy, x i , x j are the error factors, is the error range, is the correlation coefficient of the two errors, and xterm is the coupling term error of the correlated errors.

[0061] The constructed objective function is:

[0062]

[0063] Among them, is the weight value of each component in the objective function, and x(i)_change is the error amount after controlling each error factor, obtained according to the implementation difficulty of the main technologies of the simulation system.

[0064] The particle swarm optimization process is as follows:

[0065] This algorithm is initialized as a group of random particles (random solutions), flying at a certain speed in the D-dimensional search space. When each particle searches, it takes into account the historical best point it has searched and the best points of other particles in the group, and changes its position based on this.

[0066] As Figure 1 shown, a method for optimizing the error allocation of a visual simulation system according to the present invention includes the following steps:

[0067] (1): Based on the radiation energy transfer process of the simulation system, comprehensively analyze the error influencing factors, and construct a radiation energy error transfer function of the simulation system based on the error transfer theory;

[0068] (2): For the main error factors, construct an objective function for optimizing the error allocation of the simulation system;

[0069] (3): Determine the range of each error factor according to the technical feasibility of actually controlling each error factor;

[0070] (4): Design an experiment, use the error transfer function as the optimization constraint function, and use the particle swarm optimization algorithm to optimize the objective function within the adjustment range of each error factor, and output the optimal error allocation scheme of the simulation system.

[0071] As Figure 2 shown, the visual simulation energy transfer process includes a digital scene generation and a digital simulation part; among them, the calculation principle formula of the digital scene is as follows:

[0072] L(θ,λ) = {εL T + [E S cos(θ S ) + Ld](1 - ε)}τ + Lu

[0073] Among them, ε is the emissivity of the object, and L T is the energy radiated by the object itself; τ2 is the transmittance of the observation path, and ES is solar radiation, and θ S is the angle between the sun and the normal vector of the target surface; Ld is the downward atmospheric radiation, and Lu is the upward atmospheric radiation.

[0074] The physical projection principle formula is as follows:

[0075]

[0076] Among them, E(α) is the energy of the target point after physical projection, and L 黑体 is the blackbody light source; [L min , L max is the quantization range of the scene radiance, and L 数字 is the radiance calculated by the digital scene generation part; s is the area of a single pixel of the DMD; f is the total focal length of the projection optical system; τ is the total transmittance of the projection system; α is the angle between the calculated pixel and the center point of the field of view;

[0077] After the principle analysis, the main error factors and dimensions of the simulation system are summarized in Table 1:

[0078] Table 1 Main Error Factor Table

[0079]

[0080] Based on the rule of constructing the error transfer function and combining the main error factors, the radiation energy error transfer function is constructed as follows:

[0081]

[0082] Among them, L is the object energy, and error L is the error amount of the object energy.

[0083] The present invention uses a search and optimization method to realize the optimization control of the simulation system fidelity. When designing the optimization objective function, the sensitivity of each error factor and the technical cost of realizing the control error should be fully considered. Under the condition that the technical cost cannot be estimated, the objective function is designed according to the main error factors in Table 1. In order to eliminate the influence of the dimensions and value ranges of the error factors of the simulation system, first, a normalization operation is performed on each factor:

[0084]

[0085] Among them, the error amount of the error factor is the error control range, and the true estimated value of the error factor refers to the actual measurement and estimation value in the scene.

[0086] The objective function f(x) is designed as:

[0087]

[0088] Among them, is the weight value of each component in the objective function, and each variable in the objective function is defined in Table 2.

[0089] Table 2 Definition Table of Objective Function Variables

[0090]

[0091] Take the system error output by the simulation system error transfer model as the optimization constraint condition:

[0092]

[0093] As Figure 3 shown, the optimization process uses the particle swarm optimization (PSO) algorithm. This algorithm is initialized as a group of random particles (random solutions), flying at a certain speed in the D-dimensional search space. When each particle searches, it takes into account its own historically best point found and the best points of other particles in the group, and changes its position based on this.

[0094] The position of the i-th particle is expressed as: x i =(x i1 ,x i2 ,···,x iD )

[0095] The speed of the i-th particle is expressed as: v i =(v i1 ,v i2 ,···,v iD )

[0096] The historically best point experienced by the i-th particle is expressed as: 1 ≤ i ≤ m; p i =(p i1 ,p i2 ,···p iD )

[0097] The best point experienced by all particles in the group is expressed as: p g =(p g1 ,p g2 ,···,p gD )

[0098] Generally speaking, the position and speed of particles are both taken in the continuous real number space, and the position and speed of particles change according to the following equations during the iteration process:

[0099]

[0100]

[0101] Among them, ω is the inertia weight, which is used to adjust the global and local search capabilities of the algorithm; c1 and c2 are learning factors or acceleration coefficients, generally normal constants, and often take the value of 2. The learning factors enable the particles to have the ability to summarize themselves and learn from excellent individuals in the group, so as to approach their own historical optimal points and the historical optimal points within the group. ξ, η ∈ (0, 1) are pseudo-random numbers uniformly and randomly distributed within the interval (0, 1).

[0102] During the particle iteration process, the constraint condition and the objective function have the characteristic of converging in opposite directions. On the one hand, the constraint function limits the total system error within a known range. On the other hand, the continuous optimization of the objective function will increase the component error as much as possible to make the total error approach the limit value. The combination of the two gives the particles a clear flight direction, and the final optimization result is to make the fidelity of the simulation system in the constraint function greater than the set value.

[0103] The initial values and regulation ranges of the error factors are shown in Table 3:

[0104] Table 3 Initial calibration values of error factors

[0105]

[0106] In this experiment, the energy error transfer function is used to optimize the error distribution control of the simulation system, and the control target of the total system error is set to be no greater than 22%. According to the constraint formula, the optimization objective function formula, and the error initial setting and regulation range conditions in Table 3, the optimized control result of the system error distribution is as Figure 4a and Figure 4b shown.

[0107] The total system error is optimized and controlled from 31% to 21.94%. The optimized control results of the error distribution are shown in Table 4.

[0108] Table 4 Comparison table of optimized control results

[0109]

[0110] After obtaining the error amount of each error factor, the error amount is used to adjust and optimize the simulation software to make the simulation accuracy of the simulation software higher.

Claims

1. An error allocation optimization design method for a visual simulation system, characterized in that, It includes the following steps: Construct an energy error transfer function based on the object energy of the object in the urban environment; Construct an objective function based on the error factors and determine the adjustment range of each error factor; Using the particle swarm optimization algorithm, take the energy error transfer function as the constraint function, optimize the objective function within the adjustment range, and obtain the error amount of each error factor as the optimal error allocation scheme; The object energy L is: L(θ,λ) = {εL T + [E S cos(θ S ) + Ld](1 - ε)}τ + Lu where ε is the emissivity of the object, L T is the energy radiated by the object itself; τ is the transmittance of the observation path, E S is the solar radiation, θ S is the angle between the sun and the normal vector of the target surface; Ld is the downward atmospheric radiation, and Lu is the upward atmospheric radiation; When the error factors are independent of each other, that is, a change in one error does not cause a change in another error, the energy error transfer function is: where L is the energy of the object, θ is the error factor, Δθ is the error range; When the error factors are correlated, that is, a change in one error causes a change in another error, the energy error transfer function is: error 误差 = error L + xitem where xitem is the coupling term error; The objective function is: Among them, is the weight value of each component in the objective function, x(i)_change is the error amount after controlling the i-th error factor, and x(i)_error is the error calibration amount of the i-th error factor.

2. The error allocation optimization design method of a visual simulation system according to claim 1, characterized in that The coupling term error is: Among them, L is the object energy, x i , x j are error factors, is the error range, is the correlation coefficient of the two errors.

3. The error allocation optimization design method of a visual simulation system according to claim 1, characterized in that The error influencing factors include: the self-thermal radiation of the target, the emissivity of the target, the transmittance of the observation path, the incident light angle on the target surface, the upward atmospheric radiation, the downward atmospheric radiation, and the focal length.

4. An error allocation optimization design method for a visual simulation system according to claim 1, characterized in that The particle swarm optimization algorithm is specifically: Regard the error factors as particles and initialize them as a group of random particles, fly at a set speed in the D-dimensional search space. When each particle searches, when it finds the optimal solution, replace the current position with the optimal solution and continue to search until the search ends.

Images